2 hrs ago
Former OpenAI Employee Calls for Aviation-Like AI Safety
David Robinson used to work at OpenAI.
He says companies building powerful AI should be much more careful.
He thinks they should use several safety checks, like aviation and nuclear power do.
He worries that AI might behave differently outside a test.
He also says the AI industry has not clearly defined how to make AI follow human values.
Some AI leaders said they would create their own safety rules.
President Donald Trump opposes government rules for AI.
Robinson argues that careful planning could help prevent serious problems.
David Robinson, a former OpenAI employee, argued that advanced AI labs need stronger safety safeguards.
Robinson said he worked at OpenAI for three-and-a-half years and helped write safety reports for 12 advanced model launches.
He urged AI labs to adopt layered safeguards and careful planning similar to aviation or nuclear power.
Robinson warned that AI models may act differently when deployed than when tested, and said the industry has not defined or mastered alignment.
After a White House meeting, AI executives said they would pursue self-regulation, while Donald Trump opposed government regulation.
- Who
- David Robinson, a former OpenAI employee, and AI industry executives.
- What
- Robinson called for AI labs to use stronger, layered safety safeguards; executives said they would pursue self-regulation.
- Where
- The essay appeared in The Atlantic, and the executives met at the White House.
- When
- Robinson's essay was published in The Atlantic; the article also reports a White House meeting on Tuesday, without specifying a date.
- Why
- Robinson said rapid AI development and possible differences between tested and deployed behavior make stronger safety measures necessary.
Stronger safeguards and oversight
Self-regulation and limited government rules
How AI safety should be managed
Stronger safeguards and oversight
Robinson argued that AI labs should adopt redundant safeguards and careful planning modeled on aviation or nuclear power.
Self-regulation and limited government rules
AI executives said they had committed to self-regulation, including internal controls and outside analysts.
Government regulation
Stronger safeguards and oversight
The article reports calls for greater AI safety measures, but does not say Robinson specifically advocated government regulation.
Self-regulation and limited government rules
Donald Trump is vehemently against government regulation of AI and has called claims that AI is dangerous a “hoax.”
Key facts
- Former employer
- OpenAI
- Robinson's tenure
- Three-and-a-half years
- Safety reports
- Robinson said he oversaw reports for 12 advanced model product launches.
- Proposed safety model
- Layered safeguards and careful planning, comparable to aviation or nuclear power.
- Executives' stated approach
- Self-regulation, including internal controls and outside analysts.
- Trump's position
- He opposes government regulation of AI and called concerns about AI danger a “hoax.”
Quotes
David Robinson
Former OpenAI employee and AI safety researcher
“An environment where things like this can happen is no place to grow artificial minds that could be smarter than we are and that might not do what we want them to.”
NDTV
Sources
Related news
Former OpenAI Safety Employee Warns Company Culture Threatens AI Safety
OpenAI Safety Employee Quits, Says Time for Trial and Error Is Over
OpenAI Discloses Six AI Misalignment Cases, Adds Reporting Framework
OpenAI Parts Ways With Three Researchers Over Confidential Data
AI Agents’ Rogue Behavior Raises Questions About Human Control
Nvidia Launches Open Agent Safety Platform After Hugging Face Hack
Trump Approves Secret Russia Deal Amid Ukraine Conflict
Rajnath Singh Highlights Indian Navy’s Expanding Blue-Water Maritime Role
Reports Say Egypt and UAE Warned Netanyahu Before Attack
